Summary

An image in the conversation where you needed it.

It is deliberately one tool doing one thing: the prompt, the size, and optionally which model — enough to place a generated image into whatever you are already working on without opening a separate product.

What it is

A single-tool image generator over Together AI. It takes a prompt and optional dimensions, runs the request, and returns the image — with clear errors when the prompt is rejected or the API fails, rather than an empty result.

What you get

`generate_image` — takes `prompt`, with optional `width` and `height` in pixels, and an optional `model` naming the exact Together AI model; an incorrect name falls back to the default rather than failing
Generation on the Flux.1 Schnell model by default
Error handling that distinguishes a prompt validation problem from an API one

Requirements

A Together AI API key in `TOGETHER_AI_API_KEY`, set in the client config. Python 3.12 or higher, with `httpx` and `mcp`; the client runs it with `uv --directory /ABSOLUTE/PATH/TO/image-gen/ run image-gen`. Costs land on your Together AI account, so the model you name in `model` is the one you pay for.